xref: /aosp_15_r20/external/XNNPACK/src/x32-transposec/gen/4x4-multi-multi-sse2.c (revision 4bdc94577ba0e567308109d787f7fec7b531ce36)

// Auto-generated file. Do not edit!
//   Template: src/x32-transposec/sse2.c.in
//   Generator: tools/xngen
//
// Copyright 2021 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.

#include <immintrin.h>

#include <assert.h>

#include <xnnpack/common.h>
#include <xnnpack/math.h>
#include <xnnpack/transpose.h>
#include <xnnpack/unaligned.h>


void xnn_x32_transposec_ukernel__4x4_multi_multi_sse2(
    const uint32_t* input,
    uint32_t* output,
    size_t input_stride,
    size_t output_stride,
    size_t block_width,
    size_t block_height) XNN_OOB_READS
{
  assert(output_stride >= block_height * sizeof(uint32_t));
  assert(input_stride >= block_width * sizeof(uint32_t));

  const size_t tile_height = 4;
  const size_t tile_width = 4;
  const size_t tile_hbytes = tile_height * sizeof(uint32_t);
  const size_t tile_wbytes = tile_width * sizeof(uint32_t);
  const size_t input_reset = tile_wbytes - round_down_po2(block_height, tile_height) * input_stride;
  const size_t input_offset = tile_height * input_stride;
  const size_t output_reset = tile_width * output_stride - round_down_po2(block_height, 2) * sizeof(uint32_t);

  const uint32_t* i0 = input;
  const uint32_t* i1 = (const uint32_t*) ((uintptr_t) i0 + input_stride);
  const uint32_t* i2 = (const uint32_t*) ((uintptr_t) i1 + input_stride);
  const uint32_t* i3 = (const uint32_t*) ((uintptr_t) i2 + input_stride);
  uint32_t* o0 = (uint32_t*) output;
  uint32_t* o1 = (uint32_t*) ((uintptr_t) o0 + output_stride);
  uint32_t* o2 = (uint32_t*) ((uintptr_t) o1 + output_stride);
  uint32_t* o3 = (uint32_t*) ((uintptr_t) o2 + output_stride);

  do {
    if XNN_UNPREDICTABLE(block_width < 2) {
      o1 = o0;
    }
    if XNN_UNPREDICTABLE(block_width <= 2) {
      o2 = o0;
    }
    if XNN_UNPREDICTABLE(block_width < 4) {
      o3 = o0;
    }
    size_t bh = block_height;
    for (; bh >= 4; bh -= 4) {
      const __m128i v2_0 = _mm_loadu_si128((const __m128i*) i0);
      i0 = (uint32_t*) ((uintptr_t) i0 + input_offset);
      const __m128i v2_1 = _mm_loadu_si128((const __m128i*) i1);
      i1 = (uint32_t*) ((uintptr_t) i1 + input_offset);
      const __m128i v2_2 = _mm_loadu_si128((const __m128i*) i2);
      i2 = (uint32_t*) ((uintptr_t) i2 + input_offset);
      const __m128i v2_3 = _mm_loadu_si128((const __m128i*) i3);
      i3 = (uint32_t*) ((uintptr_t) i3 + input_offset);

      const __m128i v1_0 = _mm_unpacklo_epi32(v2_0, v2_1);
      const __m128i v1_1 = _mm_unpackhi_epi32(v2_0, v2_1);
      const __m128i v1_2 = _mm_unpacklo_epi32(v2_2, v2_3);
      const __m128i v1_3 = _mm_unpackhi_epi32(v2_2, v2_3);

      const __m128i v0_0 = _mm_unpacklo_epi64(v1_0, v1_2);
      const __m128i v0_1 = _mm_unpackhi_epi64(v1_0, v1_2);
      const __m128i v0_2 = _mm_unpacklo_epi64(v1_1, v1_3);
      const __m128i v0_3 = _mm_unpackhi_epi64(v1_1, v1_3);



      _mm_storeu_si128((__m128i*) o3, v0_3);
      o3 = (uint32_t*) ((uintptr_t) o3 + tile_hbytes);
      _mm_storeu_si128((__m128i*) o2, v0_2);
      o2 = (uint32_t*) ((uintptr_t) o2 + tile_hbytes);
      _mm_storeu_si128((__m128i*) o1, v0_1);
      o1 = (uint32_t*) ((uintptr_t) o1 + tile_hbytes);
      _mm_storeu_si128((__m128i*) o0, v0_0);
      o0 = (uint32_t*) ((uintptr_t) o0 + tile_hbytes);
    }
    if (bh != 0) {
      const __m128i v2_0 = _mm_loadu_si128((const __m128i*) i0);
      if XNN_UNPREDICTABLE(bh < 2) {
        i1 = i0;
      }
      const __m128i v2_1 = _mm_loadu_si128((const __m128i*) i1);
      if XNN_UNPREDICTABLE(bh <= 2) {
        i2 = i0;
      }
      const __m128i v2_2 = _mm_loadu_si128((const __m128i*) i2);
      const __m128i v2_3 = _mm_undefined_si128();

      const __m128i v1_0 = _mm_unpacklo_epi32(v2_0, v2_1);
      const __m128i v1_1 = _mm_unpackhi_epi32(v2_0, v2_1);
      const __m128i v1_2 = _mm_unpacklo_epi32(v2_2, v2_3);
      const __m128i v1_3 = _mm_unpackhi_epi32(v2_2, v2_3);

      __m128i v0_0 = _mm_unpacklo_epi64(v1_0, v1_2);
      __m128i v0_1 = _mm_unpackhi_epi64(v1_0, v1_2);
      __m128i v0_2 = _mm_unpacklo_epi64(v1_1, v1_3);
      __m128i v0_3 = _mm_unpackhi_epi64(v1_1, v1_3);



      if (bh & 2) {
        _mm_storel_epi64((__m128i*) o3, v0_3);
        o3 += 2;
        _mm_storel_epi64((__m128i*) o2, v0_2);
        o2 += 2;
        _mm_storel_epi64((__m128i*) o1, v0_1);
        o1 += 2;
        _mm_storel_epi64((__m128i*) o0, v0_0);
        o0 += 2;
        v0_0 = _mm_unpackhi_epi64(v0_0, v0_0);
        v0_1 = _mm_unpackhi_epi64(v0_1, v0_1);
        v0_2 = _mm_unpackhi_epi64(v0_2, v0_2);
        v0_3 = _mm_unpackhi_epi64(v0_3, v0_3);
      }

      if (bh & 1) {
        unaligned_store_u32(o3, (uint32_t) _mm_cvtsi128_si32(v0_3));
        unaligned_store_u32(o2, (uint32_t) _mm_cvtsi128_si32(v0_2));
        unaligned_store_u32(o1, (uint32_t) _mm_cvtsi128_si32(v0_1));
        unaligned_store_u32(o0, (uint32_t) _mm_cvtsi128_si32(v0_0));
      }
    }

    i0 = (const uint32_t*) ((uintptr_t) i0 + input_reset);
    i1 = (const uint32_t*) ((uintptr_t) i0 + input_stride);
    i2 = (const uint32_t*) ((uintptr_t) i1 + input_stride);
    i3 = (const uint32_t*) ((uintptr_t) i2 + input_stride);
    o0 = (uint32_t*) ((uintptr_t) o0 + output_reset);
    o1 = (uint32_t*) ((uintptr_t) o1 + output_reset);
    o2 = (uint32_t*) ((uintptr_t) o2 + output_reset);
    o3 = (uint32_t*) ((uintptr_t) o3 + output_reset);
    block_width = doz(block_width, tile_width);
  } while (block_width != 0);
}